What this is
This listing covers subnet 41 on the Bittensor network — Sportstensor — and the trading front end it runs on, Almanac. The subnet exists to make AI models compete on real sports predictions and to pay the ones that get them right. It is an information layer rather than a venue you sign up to deposit into: the markets it scores against are live sports markets, and what it pays out is subnet alpha, not a ticket payout.
Three names turn up interchangeably in coverage and in the block explorers, so it is worth separating them. Sportstensor is the subnet (SN41 on mainnet, 172 on testnet). Almanac is the product miners actually trade through, at almanac.market. Bittensor is the parent network that supplies the wallets, UIDs and emissions the whole thing runs on. The explorer taostats lists SN41 under the name Almanac, which is the front end rather than the subnet itself.
Why it belongs in a prediction-market directory
Most entries here are venues, dashboards, data feeds or development shops. SN41 is a different shape: it is the forecasting layer that could sit underneath the others. Miners do not publish a probability and hope to be believed — they express it by trading, and the subnet pays them on trading performance rather than on self-reported accuracy. That makes it the piece of the stack that produces and ranks signals, not the piece that settles bets.
How a miner earns
A miner trades on Almanac. Almanac routes those orders to the Polymarket central limit order book through the miner’s own proxy wallet, so the positions are the miner’s and the execution is on a real book. Each trade then becomes a scored prediction inside the subnet’s incentive mechanism, which weighs accuracy, ROI, timing and informational value. The scoring is deliberately model-agnostic: a hand-traded view, a statistical model and a fully automated system are judged on the same terms, which is the whole point of the design.
How the scoring works
Validators run a two-phase optimisation over a rolling 30-day window, divided into 24-hour epochs. For each epoch they compute ROI (profit divided by trading volume) and qualified volume (volume from winning trades, after fees), apply eligibility gates — a minimum ROI, a minimum volume and a build-up period — then optimise twice: phase one maximises the qualified volume that can be funded inside the fixed budget, and phase two redistributes payouts toward higher-ROI traders while still holding the volume targets. The result is converted into token weights and paid as daily alpha emissions, with diversity caps so no single trader dominates, volume decay so recent activity counts for more, and ramp constraints so allocations do not jump suddenly.
Scoring runs hourly in the background against the previous epoch while weights for the current one are updated. Validators publish the final weights on-chain, and those weights are what decide the next epoch’s emissions.
What it takes to take part
You need a Bittensor wallet with a UID registered on SN41, an Almanac account linked to Polymarket and to the coldkey behind that UID, and Python 3.10+ to run the miner script. Miners who would rather not run code can trade directly in the Almanac app once the account is linked. Running a validator needs a validator permit on the subnet — stake, in other words.
Caveats
No performance claims on this page were verified by Bitnxt: the mechanism is described from the subnet’s own repository and docs, not from a backtest we ran. Almanac was in beta at the time of writing, so interfaces and onboarding steps can change. Emissions are paid in subnet alpha, whose realised value depends on that token’s liquidity and on emission rates that the network itself adjusts — a high score is not a cash figure. Registration on a full subnet costs TAO. And note the link on this listing points at bittensor.com, the parent network, rather than the subnet’s own front end; the product lives at almanac.market and the code at github.com/sportstensor/sn41.





































